BigO(Bench) - Space Complexity Ranking: leaderboard

Each model, zero-shot, must write an optimised solution in a requested space-complexity class; the score is the percentile of the best measured memory coefficient of 20 attempts among human solutions of that class (0 when no attempt passes the tests); each model ran on the 308 problems (636 problem-class pairs) of the test set. These are the original protocol's numbers (paper v1/v2, March 2025, problem statements with their input bounds); the paper's 2026 v3 drops the bounds and re-measures, and is not this board. Llama 3.1 405B, which generated the input parsers that decide which problems enter the test set, is not ranked. Unlike HumanEval-style correctness boards it scores whether code meets a stated algorithmic complexity.

Metric: coeffFull (%). Source: facebookresearch.github.io. Saturation forecast: Estimated already saturated. 16 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Llama 3.1 Nemotron Ultra 253B v145.4#1014
2DeepSeek R1 Distill Llama 70B41.6#1669
3QwQ-32B41.1#908
4DeepSeek R1 Distill Qwen 32B40.1#1806
5DeepSeek V3 (0324)39.7#722
6Llama 4 Scout Instruct34.5#1575
7Llama 3.3 70B Instruct32.8#1387
8DeepSeek V327.2#734
9Llama 3.1 70B Instruct26.9#1577
10DeepSeek-Coder-V2-Instruct-072426.8#1103
11Codestral-22B-v0.125.2#2680
12openhands-lm-32B-v0.122.4#2237
13Llama 4 Maverick Instruct FP821.8#1022
14Qwen 2.5 Coder 32B Instruct20.5#1597
15Gemma 3 27B (IT)17.5#1339

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=bigo-bench-space-complexity-ranking · How It Works · Data refreshed daily, snapshot 2026-11-17.